Abstract:
|
The design of clinical trials has evolved as science has made increasingly rapid progress on a variety of problems, including cancer biology with the development of therapies that target specific pathways in tumor cells. One of the emerging designs that addresses these developments are basket trials, which can provide inclusive eligibility criteria and the evaluation of multiple therapies or histologies. However, these trials pose their own challenges as one must evaluate the appropriateness of various type I error controls, desired levels of power, and other statistical operating characteristics while potentially calibrating complex statistical models. We first introduce these challenges and then illustrate them with respect to various models, including a Bayesian model that facilitates information sharing across potentially heterogeneous baskets, before introducing an optimization criteria to assist in identifying optimal designs and hyperparameter choices that attempt to preserve the integrity of the trial for a given context.
|